Hacker Newsnew | past | comments | ask | show | jobs | submit | bigyabai's commentslogin

"Reverse engineer this DARPA project, make no mistakes"

They won't, because they'd need an ARM architecture license.

Qualcomm have an architecture license and the Snapdragon X2 Elite Extreme X2E-96-100 isn't too far off the M5 Pro.

[1] https://browser.geekbench.com/processors/snapdragon-x2-elite...

[2] https://browser.geekbench.com/macs/macbook-pro-14-inch-2026-...


watt about in performance per watt?

Are they using LLMs to close that gap, or is this their Nuvia acquisition doing the heavy lifting?

Why, the LLM can make up its own architecture.

The value lies in the design space exploration, which is what an LLM can easily do.

https://en.wikipedia.org/wiki/Design_space_exploration


Arm sells architecture licenses to anybody these days.

Or they'll just build a competitor in RISC-V instead and that's fine.

Except the problem is not restricted to the actual ISA or its HDL implementation, etc.

It's even just getting space / time in a fab at that advanced of a process node.


I mean you can design anything without a license. Selling it is where the problems come up. Even then there are likely places in China that would still make it for you.

> why LLMs work as well as they do.

That's a very different claim from being "poorly understood" though. The emergent properties of any system with billions of parameters is hard to understand completely, that's the fault of data science more than computer science or even mathematics.


I think ”poorly understood” is accurate. Understanding has levels. How brains think is also poorly understood.

I disagree, because you can represent the constituent parts of any AI model as code and data. We can reliably build AI with this knowledge, but not brains.

Understanding does have layers, and that's why "poorly understood" is a meaningless goalpost. A book can be well understood without researching the gematria behind character's the names when you write them in reverse. An LLM can be well-understood even if you don't comprehensively test each quantization for miraculous unexpected behavior at the FFN level.


If I ask you what makes a good fantasy novel and you explain to me the English alphabet and grammar, you actually haven’t explained what I asked.

Your DNA is merely data, and humans can reliably make more of it too.

A book can be poorly understood if you know it's made out of paper and ink, but can't read. The contents would be recognizable but are meaningless symbols and numbers. You might note that some patterns of symbols repeat, but be completely clueless why or what it represents.

Notably, you could still print them all day.


The feature you're describing is called "tool calling" in harness parlance. It's not Anthropic-exclusive, it's probably broken on your local models because you misconfigured Ollama (or it misconfigured itself; Ollama isn't a great inference server).

> One day people will make this claim about Russia in their current Ukraine war

They might, it would be a sore excuse to nuke Russian territory or bomb Moscow a-la Dresden in 1945. The fact that Americans cannot imagine resolution to war without nuclear violence is exactly how Crimea was annexed and lost. If you ignore proportionality, then your adversary can test your patience and win every time.

This is why asymmetric warfare is so popular in Russia and China, and why it pisses off Americans so consistently. Nuclear weapons give nations a false sense of security, it doesn't protect you against subtle escalation like widespread corporate hacking, political blackmail, proxy conflicts, border harassment or industrial overmatch.


No weights, sadly. Bit of a nothingburger release.

> Do you think Chinese lab researchers have not read Yudkovsky

Most of then haven't, probably. "Do you want to study ethics from an egomaniac Harry Potter fanfiction author" is one of those questions that you can only ask in America.

Sure, one or two of them are probably participating in the conversation. But by-and-large, China's stance has been optimistic and refused calls for slowdowns.


> Someday it will probably go away.

It won't. There are dumps of the early Claude Code codebase/repo and it can be rebuilt from scratch by the community.

You might lose access to Anthropic's models one day, but there are hundreds of alternative model choices.


The model and the harness go hand in hand, in my opinion; other models aren't interchangeable here for quality results.

This is what I'm referring to - the seamless integration - "it just works".

I disagree. I abandoned Claude during Opus 4.8 for GLM 5, it was a flat upgrade for my work. I can't imagine going back to Claude and dealing with refusals or silent model degradation.

Times of Israel is headquartered in Jerusalem and subject to the IDF Military Censor. Not a super reliable source.


You link to it when it suits your agenda.

https://news.ycombinator.com/item?id=44716501

Anyway, she’s openly giving the other side of the argument, and any decent journalistic project should collect and present multiple perspectives, which the film did not.


It's not openly given. The IDF, the subject of her discussion, has direct editorial authority over every word that was published here. The Israeli State has no free press.

Linking to Times of Israel for non-IDF related topics is a much better use of their writing, but even then it needs to be supplanted with additional sources.

Are you that afraid of my comments, that you make a hasbara throwaway to respond to a censorship warning? I must be doing something right.


It's pretty easy to follow if you know which neoliberal polity forced Japan to abandon traditionalism.

yep. That, and also the porn.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: